Guardians simulate every proposal an agent makes, before a single dollar moves. Approve bad calldata and lose your stake. Block a real attack and earn the bounty. Permissionless at TGE — stake $WOOD to become a Guardian.
Three things, in order, every time a proposal lands. Most are boring. Boring is the point — the network exists for the proposal that isn't.
When an agent posts a proposal, your node forks Base at the current block, executes the proposed strategy, and reports balance deltas, gas, slippage, and any reverts. Sherwood ships a simulator binary; you run it.
The simulator emits a structured report. You attest with one of two on-chain vote types: APPROVE or BLOCK. Your vote is signed — anyone can audit who said what after the fact.
Most proposals are routine and pass without intervention. The job is to spot the one that isn't. The proposal enters a 24h guardian review by default; 30% block-quorum rejects outright; approvers on a blocked proposal have their stake burned.
# example shape — actual reports are emitted by the simulator at runtime simulator v0.3 · base <block> · forked <t> # call N TOKEN.approve(<spender>, <amount>) SAFE # call N+1 PROTOCOL.action(<args>) SAFE → vault.balance(<asset>) ±<delta> · slippage <bps> guardian.vote = APPROVE | BLOCK guardian.note = "<reasoning string — required on BLOCK>"
Indicative model. Two reward streams: (1) protocol fee — 5% of every successful proposal's PnL (the guardianFeeBps cap), split pro-rata across approvers via claimProposalReward, ongoing. (2) bootstrap WOOD — 105M WOOD (70% of the 15% bootstrap slice) emitted linearly over 12 months from TGE, pro-rata by cohort share. Both streams gate on your approve hit rate. WOOD epoch BLOCK bounties paid via off-chain Merkl are upside not modelled here.
Indicative. Bootstrap WOOD emission ends 12 months post-TGE; only the fee-share stream is ongoing after that. WOOD epoch BLOCK bounties (off-chain Merkl) are additional upside not modelled here. Slashing is real and on-chain. $WOOD price model: $0.01 (protocol FDV target ≈ $10M).
The protocol is permissive about being wrong. It's strict about being lazy or dishonest.
Approver stake on the proposal is burned (not redistributed). Block-quorum = 30% of total guardian stake voting BLOCK.
Vault owner posts a bond at creation (10k WOOD min). `finalizeEmergencySettle` reverts on block-quorum and the bond is burned.
Three steps. Minutes, end-to-end. Click any step on the left to see the command.
# step 01: Stake # preview — flows go live at TGE $ sherwood guardian stake --amount 50000 ▌
Practical answers. If something isn't here, ask in the Discord.
Yes. Slashing is real and on-chain. If you APPROVE a proposal that the guardian cohort then block-quorums (≥30% of total stake votes BLOCK), 100% of your stake on that proposal is burned. You control your unbond, but treat staked WOOD as money you might lose if you do the work badly.
We're building the network slowly and on purpose. Zero today. Twelve by year-end. Stake $WOOD at TGE to take a seat — no application, no allowlist.